R

Sw

Ma

b

c

a

ARR2A

KSRYWI

1

scW2dii2iop

–

h0

Agricultural Water Management 195 (2018) 37–46

Contents lists available at ScienceDirect

Agricultural Water Management

journa l homepage: www.e lsev ier .com/ locate /agwat

esearch paper

patio-temporal salinity dynamics and yield response of rice inater-seeded rice fields

athias Marcos a,∗, Hussain Sharifi b, Stephen R. Grattan c, Bruce A. Linquist a

Department of Plant Sciences, University of California – Davis, Davis, CA, 95616, USAThe Wonderful Company, Los Angeles, CA, 90064, USADepartment of Land, Air and Water Resources, University of California – Davis, Davis, CA, 95616, USA

r t i c l e i n f o

rticle history:eceived 23 December 2016eceived in revised form3 September 2017ccepted 26 September 2017

eywords:alinity dynamicsiceield thresholdater quality modeling

rrigation water quality

a b s t r a c t

The scarcity of high quality irrigation water is a global issue facing rice growers, forcing many to adoptwater management systems that may result in increased salinity and yield reductions. While salt con-centrations in field water have been shown to vary depending on water management, the distributionand build-up patterns of dissolved salts are unclear. This study was conducted to elucidate the withinfield spatial and temporal salinity dynamics in water-seeded rice cropping systems, and to assess currentsalinity thresholds for rice yield reduction. In this two-year study, water and soil salinity concentrationsof eleven field sites were monitored weekly, with three sampling points being established in the top, mid-dle and bottom basins of each field. There was a consistent spatio-temporal water salinity pattern amongall fields: the maximum water salinity within a field occurred during week 2 to week 7 after planting,and was greatest farther from the irrigation inlet and where soil salinity was high. A model developed topredict water salinity within a field indicates that, averaged over an entire growing season, the positionwithin a field contributed to 82% of the variation explained by the model, while preseason soil salinity

contributed to 18%. Importantly, field water salinity was determined to be the most sensitive salinitymetric for rice yield, as preseason soil salinity was a poor predictor of yield loss. The threshold field watersalinity concentration was estimated at 0.88 dS m−1, lower than the previous report of 1.9 dS m−1. Theseresults illustrate the ability to predict water salinity in a rice field with few parameters, while highlightingthe importance of field water salinity as the main salinity metric for rice cropping systems.

© 2017 Elsevier B.V. All rights reserved.

. Introduction

High quality water available for irrigated agriculture is currentlycarce and is expected to become less available due to climatehange and population growth (Hanak and Lund, 2012; Fraiture and

ichelns, 2010; Mirchi et al., 2013; Schewe et al., 2014; Wallace000). This will result in increased use of marginal water andecreased drainage, thereby resulting in increased secondary salin-

zation (Connor et al., 2012; Molden et al., 2010). Rice, a globallymportant staple crop, is the most salt-sensitive cereal (Grieve et al.,012; Munns and Tester, 2008). Additionally, when grown under

rrigation, rice requires two to three times more water input thanther cereals (Bouman et al., 2007; Kijne 2006). The current androjected decline in the quantity and quality of water for rice pro-

∗ Corresponding author at: Department of Plant Sciences, University of California Davis, Davis, CA, 95616, USA.

E-mail address: mmarcos@ucdavis.edu (M. Marcos).

ttps://doi.org/10.1016/j.agwat.2017.09.016378-3774/© 2017 Elsevier B.V. All rights reserved.

duction, prompts the need to investigate salinity in rice croppingsystems to avoid yield reductions.

In California, rice is the top agricultural water user based onapplication rate per hectare (USDA, 2013). Rice fields in Cal-ifornia, which are typically divided into several hydrologicallyconnected basins, are continuously flooded throughout the grow-ing season, with irrigation water entering the topmost basin andcascading through to the bottommost basin. The primary watermanagement system in California is a conventional flow-throughsystem (hereafter referred to as “conventional-drainage”). Underconventional-drainage, tailwater discharges to a drainage ditch formuch of the growing season; the exception being during waterholding periods after pesticide applications, whereby tailwaterdrainage is temporary halted to allow for pesticide degradationin the field (Hill et al., 1991). Under conventional-drainage, theamount of tailwater drainage can be as high as 7.6 ML ha−1 (Hill

et al., 2006) and 39% of the total water applied to a field (Linquistet al., 2015). Tailwater drainage helps remove salts that accumu-late in the field water (Scardaci et al., 2002), thereby preventing

https://doi.org/10.1016/j.agwat.2017.09.016http://www.sciencedirect.com/science/journal/03783774http://www.elsevier.com/locate/agwathttp://crossmark.crossref.org/dialog/?doi=10.1016/j.agwat.2017.09.016&domain=pdfmailto:mmarcos@ucdavis.eduhttps://doi.org/10.1016/j.agwat.2017.09.016

3 ater Management 195 (2018) 37–46

tRagmtWfiiwwahl

p1ectbgtoswlits(ieisat1thfio

tm21ds1dodwmefiliddesfi

8 M. Marcos et al. / Agricultural W

he long term accumulation of salt in the soil (Lekakis et al., 2015).ecent drought conditions in California have reduced the watervailable for rice production (Howitt et al., 2015), forcing manyrowers to reduce the amount of water applied to a field. A com-on method to reduce the water applied to a field is to eliminate

he tailwater drainage (hereafter referred to as “zero-drainage”).hile zero-drainage can greatly reduce the total water applied to a

eld, it is likely to increase the salinity concentration, particularlyn bottom basins. Nevertheless, the projected future of decreased

ater available for rice production in California (Hill et al., 2006)ill likely increase the practice of zero-drainage. This, along with

n increased reliance on groundwater for irrigation, which has aigher salt concentration than surface water (Grattan 2002), may

ead to high salinity in rice fields and reductions in yield.Crop yield response to salinity is most often displayed using a

iecewise linear model (Ayers and Westcot 1985; Maas and Grattan999; Maas and Hoffman 1977), where the first segment is a tol-rance plateau with a slope of zero, and the second segment is aoncentration dependent line with a negative slope (Supplemen-ary Fig. 1). The threshold salinity concentration, the concentrationeyond which crop yields decline, is of utmost concern to bothrowers and regulators. In flooded rice systems, a confounding fac-or is the need to account for both field water salinity (i.e. salinityf the ponded water) and soil salinity. Traditionally, field water andoil salinity thresholds for rice have been developed from studiesith limited observations over a wide salinity range, resulting in

arge distances between salinity treatments and a large uncertaintyn threshold estimates (Grieve et al., 2012). Additionally, salinityhreshold studies occur under steady-state conditions, wherebyalinity stress is kept constant throughout the growing seasonMaas and Grattan 1999). Conversely, field water and soil salin-ty in commercial rice fields lack temporal uniformity (Scardacit al., 1996; Scardaci et al., 2002), and the range of observed salin-ty concentrations is narrower and lower than in salinity thresholdtudies. The lack of relatedness between field and study conditions,nd the uncertainty of threshold estimates, has led some to ques-ion the applicability of current thresholds (Kijne 2006; Shalhevet994), especially in rice, as yield loss has been reported belowhe threshold value (Simmonds et al., 2013). These discrepanciesave increased the focus on developing thresholds under realisticeld conditions (Kijne, 2006), thereby increasing the applicabilityf threshold estimates.

Elucidating the spatial and temporal salinity dynamics is vitalo ameliorate salinity stress in rice fields, especially since rice is

ore sensitive to salinity from tillering to flowering (Castillo et al.,007; Heenan et al., 1988; Fraga et al., 2010; Pearson and Bernstein959; Zeng et al., 2001). Additionally, there have been reports ofecreased stand establishment in commercial rice fields due to highalinity early in the season (Scardaci et al., 2002; Shannon et al.,998). Previous studies in commercial fields under conventional-rainage, have found that water salinity increased in bottom basinsf fields (Scardaci et al., 2002 & Simmonds et al., 2013), likelyue to evapo-concentration. Simmonds et al. (2013) also foundater salinity to be higher in areas of the field away from the pri-ary water flow path (i.e. low flow areas). Results from Simmonds

t al. (2013) and Scardaci et al. (2002), suggest that location in aeld, relative to the irrigation inlet and primary water flow path,

argely determines the field water salinity concentration; though,t is unclear how applicable these results are in fields under zero-rainage. A complete understanding of the spatio-temporal salinityynamics in rice fields, however, is essential to being able to prop-rly manage salinity in rice fields. Therefore, the objectives of this

tudy were to: 1) reassess salinity thresholds using commercial riceelds, 2) quantify spatial and temporal variation of water and soil

Fig. 1. A map illustrating the location of field sites for the study. Field names referto the first letter of the counties they are in.

salinity in commercial rice fields and 3) develop a model to predictwater salinity in fields under zero-drainage.

2. Materials and methods

2.1. Study area and site descriptions

This experiment was carried out in commercial rice fieldsthroughout the Upper Sacramento Valley of California (Fig. 1) dur-ing the 2014 and 2015 rice growing seasons. Eleven field sites withzero tailwater drainage or conventional tailwater drainage, and awide range of soil salinity, were chosen for this study. There wereseven field sites in Colusa county (C1, C2, C3, C4, C5, C6, and C7), twoin Glenn county, (G1 and G2), one in Butte county, (B1), and one inYuba county, (Y1). This region has a Mediterranean climate, char-acterized by warm, dry conditions during the rice growing season.The mean air temperature during the 2014 and 2015 growing sea-sons was 22.7 ◦C, while the mean precipitation during the growingseasons was 15 mm (CIMIS, 2016). All field sites have fine-texturedsoils with minimal slope, which is typical for rice fields in the region.Soil taxonomic classifications, soil characteristics, irrigation prac-tices and variety sown are shown in Table 1. Field-specific pesticideregimens were employed to combat weeds and insect pests.

2.2. Experimental design

Each field site contained 9 plots (2 m × 2m) that were splitbetween the top (A), middle (B) and bottom(C) basins (Fig. 2).Within each basin, three sampling plots were established and num-bered (1, 2, 3) based on their proximity to the primary water flowpath (plot 1 being closest and plot 3 being farthest from the pri-mary water flow path). For basins B and C, if water flowed downfrom both sides of the field, then plot 3 was in the middle of thebasin (as shown in Fig. 2). If water flowed from only one side ofthe field, then plots in the B and C basins were spaced similar tothe A basin. All plots were established 15 m in from the edge of thefield to avoid border effects. Fig. 2 is a representation of the plotdesign within a field; however, field sites varied in dimension andnumber of basins. The position of each plot was determined in eachfield, with the position within a field being considered as the com-bination of the longitudinal distance down the field and the lateraldistance across a basin (as shown in Fig. 2).

2.3. Sampling and measurements

After spring land preparation, but before fertilizer applicationand flooding, soil samples were taken from each plot at a depth of

M. Marcos et al. / Agricultural Water Management 195 (2018) 37–46 39

Tab

le

1O

verv

iew

of

site

soil

char

acte

rist

ics

and

wat

er

man

agem

ent.

Fiel

d

Soil

Taxo

nom

icC

lass

ifica

tion

Fiel

d

Size

(ha)

Nu

mbe

r

ofB

asin

sSo

il

Text

ure

SOC

(%)

pH

Fiel

d

ECe

(dS

m−1

)SA

R

Tail

wat

erD

rain

age

Cu

ltiv

ar

San

d

(%)

Silt

(%)

Cla

y

(%)

C1

Fin

e,

smec

titi

c,

ther

mic

Sod

ic

End

oqu

erts

44

7

10

57

33

2.15

6.26

1.13

4.60

Zero

M-2

06

C2

“

44

7

10

57

33

2.26

6.35

0.99

4.27

Zero

M-2

06C

3

“

64

6

8

56

36

2.52

6.05

0.85

3.18

Zero

M-2

06C

4

“

32

3

12

28

60

2.04

7.46

5.88

8.31

Con

ven

tion

al

M-2

06C

5

“

17

2

12

27

61

2.22

6.56

5.69

7.93

Con

ven

tion

al

M-2

02C

6

“

36

3

12

27

61

1.54

7.09

6.25

8.37

Zero

M-2

06C

7

“

21

6

21

45

34

1.79

6.93

4.49

8.43

Con

ven

tion

al

Cal

hik

ari-

202

G1

Fin

e,th

erm

ic

Typ

icC

alci

aqu

olls

45

9

20

40

40

2.20

5.76

0.50

1.17

Zero

M-2

06

G2

“

31

4

13

44

43

2.10

5.76

0.59

1.72

Zero

M-2

08B

1

Fin

e,

smec

titi

c,

ther

mic

Xer

ic

Epia

quer

ts

and

Du

raqu

erts

24

3

31

23

46

1.90

4.94

0.33

0.91

Zero

M-1

05

Y1

Fin

e,

mix

ed

ther

mic

Abr

up

tic

Du

rixe

ralf

s30

3

43

37

20

1.19

4.59

0.29

0.88

Zero

M-2

06

“:

Sam

e

as

abov

e;

SOC

:

soil

orga

nic

carb

on;

SAR

:

sod

ium

adso

rpti

on

rati

o;

ECe:

pre

seas

on

soil

satu

rate

d

pas

te

sali

nit

y.

Fig. 2. A map illustrating the experimental set up within a field. The solid lineand arrow on the left represents the longitudinal distance down the field, whilethe dashed lines and arrows represent the lateral distance across a basin. Plots arerepresented by a letter & number combination. A plots location in a field is a combi-nation of the longitudinal and lateral distance from the irrigation inlet. Water salinitymeasurements were collected at all plots. Soil solution salinity measurements were

collected at circled plots.

0–15 cm. Soils were air dried, then ground to pass through a 2-mmsieve. For each field site, equally weighted portions of the soil sam-ples were mixed to generate a composite field sample, from which,general field soil property data were determined (Table 1). Soil pH,soil texture, sodium adsorption ratio, and soil organic carbon weredetermined from the composite field samples using standard meth-ods described in the Soil Survey Laboratory Methods Manual (Burt,2014).

Four types of salinity measurements, all based on electrical con-ductivity (EC), were made in this study using a calibrated OaktonCON 450 or an Oakton CON 400 (Vernon, IL, USA) standardizedto 25 ◦C. Throughout the growing season, weekly water salinitymeasurements were made at the irrigation inlet (ECiw), and at allplot locations (ECfw). To obtain the measurements, the EC probewas carefully lowered into the water and kept submerged untila stable reading was reached. Soil samples taken from each plotprior to flooding were used to determine the preseason soil salinity(ECe) using the saturated paste technique (Burt, 2014). Additionally,weekly in-season soil solution salinity (ECss) measurements weremade, at the same time as ECfw measurements, at A1, B2, and C3plots (Fig. 2). In order to obtain ECss measurements, before flood-ing, soil pore-water samplers (Rhizon MOM 10 cm, RhizosphereResearch Products, Wageningen, The Netherlands) were installed10 cm below the soil surface. Before obtaining ECss measurements,the hose of the pore-water sampler was purged by extracting the

first 10 mL of solution into evacuated vials. Soil solution was thenextracted into evacuated 40 mL vials, and ECss was measured in situ.

40 M. Marcos et al. / Agricultural Water Management 195 (2018) 37–46

Fig. 3. Temporal pattern of average relative water EC (field water salinity relative to irrigshown on the right axis. Water depth in other basins has a similar pattern (data not show

Fig. 4. The relationship between season-average field water salinity (ECfw) andseason-average soil solution salinity(ECss).

FsT

Fa

2raagm

2

ad(a

ig. 5. The relationship between yield (kg ha−1) and season-average field wateralinity (ECfw). Yield is negatively correlated with field water EC above 0.88 dS m−1.his figure only includes the significant fixed effects.

ield water depth measurements were taken at the same locationsnd at the same time as the ECss measurements.

Upon physiological maturity (grain moisture content below8%) yield was quantified at all plot locations. A 1-m2 undisturbedepresentative area within each plot was hand harvested and then

sub-sample was oven-dried at 70 ◦C until a constant weight waschieved. Samples were then processed to determine total above-round biomass and grain yield. Grain yield was corrected to 14%oisture content.

.4. Data analysis

R statistical software (R Core Team, 2015) was used for data

nalysis and visualization. The following packages were used forata visualization: ‘ggplot2′ (Wickham and Chang 2016), ‘gtable’Wickham, 2016), ‘maps’ (Becker et al., 2016), ‘maptools’ (Bivandnd Lewin-Koh 2016), and ‘raster’ (Hijmans, 2015).

ation water salinity) at A1, B2, and C3 plots. Average water depth in the C3 plot isn). Error bars represent standard error.

For comparisons of soil and water salinity between the top (A1)and bottom (C3) of a field, a one sample t-test was performed.Results are reported as a percent increase from the top to the bot-tom basin.

Salinity yield threshold analysis was conducted by determiningthe convergence point of two linear regression lines seeking to min-imize overall deviance. To account for potential difference in yieldat the field scale, a mixed-effects modeling approach was employedusing the ‘lme4′ (Bates et al., 2016) package. In this approach, fieldsite was a random effect and was allowed to have its own intercept(i.e. yield potential).

While a water salinity threshold was determined, this study wasunable to identify a threshold for soil salinity. Nevertheless, thepotential effect of soil salinity on yield was assessed using the ECeof plot locations above the water salinity threshold. The relativeimportance of soil salinity on yield was evaluated by comparingthree mixed-effects models (1–3) developed using the ‘lme4′ (Bateset al., 2016) package:

Yij = aj + B1 ∗ xij + eij (1)

Yij = aj + B2 ∗ zij + eij (2)

Yij = aj + B1 ∗ xij + B2 ∗ zij + eij (3)

where Yij is the yield at plot i in field j, aj is the intercept for fieldj, xij is the water salinity at plot i in field j, zij is the soil salinityat plot i in field j, B1 is the water salinity coefficient, B2 is the soilsalinity coefficient, and eij is the error term for plot i in field j. Therelative importance of field water and soil salinity to yield loss wasassessed by comparing the Akaike’s Information Criterion adjustedfor a small sample size (AICc), marginal R2 values of the models,and the significance of the model coefficients. The ‘MuMIn’ (Bartoń,2016) package was used to determine AICc values of the models,with the lowest AICc value indicating the best model (Burnham andAnderson 2002). The ‘MuMIn’ (Bartoń, 2016) package was also usedto determine marginal R2 values for the mixed-effects models; themethod used follows the procedure described by Nakagawa andSchielzeth (2013), where marginal R2 is the proportion of variationexplained by the fixed factors. The ‘lmerTest’ package (Kuznetsovaet al., 2016) was used to determine p values for the mixed-effectsmodels.

As field salinity concentrations varied based on field locationand source of irrigation water, comparisons of field water salinityamong field sites were made relative to the irrigation water salinity.

During the season, the C2 site began receiving water from multipleirrigation inlets and could not be used for this analysis. Using fieldsunder zero-drainage, a predictive model for season-average rela-tive field water salinity was developed based on a priori knowledge

ater Management 195 (2018) 37–46 41

ow

S

widubiawaddacdtie2uTfilz

3

3

6ssewtrgsfbd

3d

wtaco2(lEwaE

irri

gati

on

wat

er

sali

nit

y

(EC

iw),

fiel

d

wat

er

sali

nit

y

(EC

fw),

and

soil

solu

tion

sali

nit

y

(EC

ss)

con

cen

trat

ion

s

obse

rved

in

the

stu

dy.

on

Wat

er

EC

Fiel

d

Wat

er

EC

Soil

Solu

tion

EC

on

Sou

rce

Ave

rage

ECiw

(dS

m−1

)EC

iwra

nge

(dS

m−1

)Ave

rage

ECfw

(dS

m−1

)Min

ECfw

(dS

m−1

)/Pl

ot/W

eekM

ax

ECfw

(dS

m−1

)/Pl

ot/W

eekA

vera

ge

ECss

(dS

m−1

)Min

ECss

(dS

m−1

)/Pl

ot/W

eekM

ax

ECss

(dS

m−1

)/Pl

ot/W

eek

0.47

0.31

–0.6

4

0.62

0.33

/A1/

W10

1.64

/C2/

W4

1.65

0.43

/A1/

W2

3.69

/C3/

W16

d

&

Surf

ace0

.45

0.29

–0.7

40.

76

0.25

/A1/

W3

2.10

/C2/

W5

1.79

0.43

/A1/

W8

3.51

/C3/

W6

0.62

0.31

–0.7

5

0.75

0.33

/A1/

W9

2.45

/B3/

W3

1.28

0.73

/A1/

W1

2.30

/C3/

W4

d

&

Surf

ace1

.11

0.42

–1.7

0

1.46

0.49

/C1/

W2

6.06

/C3/

W6

7.55

4.55

/A1/

W14

11.3

5/C

3/W

2

0.54

0.39

–0.6

6

0.82

0.43

/A1/

W1

4.19

/C3/

W7

5.70

1.69

/A1/

W7

12.2

8/C

3/W

1

0.71

0.41

–1.3

8

0.99

0.40

/A1/

W3

2.08

/A3/

W2

7.81

4.35

/A1/

W16

11.9

5/C

3/W

5d

1.81

1.12

–2.9

2

2.10

1.11

/A1/

W14

4.44

/C3/

W7

3.49

1.89

/A1/

W15

5.35

/C3/

W7

0.30

0.24

–0.4

3

0.37

0.19

/A3/

W5

0.72

/C3/

W3

0.84

0.31

/A1/

W13

1.56

/B2/

W3

0.28

0.18

–0.4

0

0.35

0.17

/A3/

W11

0.80

/C3/

W13

1.06

0.32

/A1/

W4

2.81

/C3/

W13

0.11

0.08

–0.2

4

0.15

0.08

/C3/

W7

0.36

/C1/

W10

0.75

0.11

/B2/

W7

1.36

/C3/

W16

0.09

0.08

–0.1

3

0.12

0.06

/B1/

W7

0.39

/C1/

W4

0.90

0.08

/A1/

W6

2.17

/C3/

W2

con

ven

tion

al

tail

wat

er

dra

inag

e;

ECiw

:

irri

gati

on

wat

er

sali

nit

y;

ECfw

:

fiel

d

wat

er

sali

nit

y;

ECss

:

soil

solu

tion

sali

nit

y.

M. Marcos et al. / Agricultural W

f potential parameters that may influence water salinity: positionithin a field and soil salinity. One superior model emerged (4):

ij = a + B3 ∗ wij + B4 ∗ uij + B5 ∗ qij + eij (4)

here Sij is the salinity of plot i in field j relative to the salinity of therrigation water into field j, a is the intercept, wij is the longitudinalistance down the field from the irrigation inlet for plot i in field j,ij is the lateral distance across a basin from the irrigation inlet orasin inlet for plot i in field j, qij is the preseason soil salinity of plot

in field j, eij is the error term for plot i in field j, and B3, B4, and B5re model coefficients. Parameters from this season-average modelere used to develop early and late-season models to assess vari-

ble contributions throughout the growing season. Early-season isefined as week 1 to week 7 after planting, while late-season isefined as week 8 to week 15 after planting. The ‘relaimpo’ pack-ge (Groemping and Matthias, 2013) was used to assess variableontributions to the models; this was done by R2 partitioning, asone by Lindeman et al. (1980). Model performance was assessedhrough leave-one-field-out (LOFO) cross-validation, as employedn other field-based agricultural studies (Pike et al., 2009; Scudierot al., 2015; Stevens et al., 2012), using the ‘caret’ package (Kuhn016). In this LOFO cross-validation approach, a model is developedsing data from six of the seven usable fields under zero-drainage.he model is then used to generate predictions for the seventheld. This procedure is repeated seven times so that each field is

eft out once. Finally, after model validation, data from all sevenero-drainage fields were used to develop the final models.

. Results

.1. Site salinity conditions

Field average preseason soil salinity (ECe) ranged from 0.29 to.25 dS m−1 (Table 1). In general, fields with lower preseason soilalinity were under zero-drainage, while fields with higher pre-eason soil salinity, were under conventional-drainage, with thexception of C6, which was under zero-drainage. The irrigationater salinity (ECiw) of fields under zero-drainage ranged from 0.08

o 1.38 dS m−1, while ECiw of fields under conventional-drainageanged from 0.39 to 2.92 dS m−1 (Table 2). Fields supplied withroundwater had a higher ECiw than fields supplied solely withurface water. Additionally, the ECiw of fields supplied with sur-ace water, which is under irrigation district control, varied largelyased on the proportion of recycled water allowed by the irrigationistrict.

.2. Spatial and temporal field water and soil solution salinityynamics

Fields differed in size and shape, leading to variation in fieldater salinity (ECfw) and soil solution salinity (ECss) build-up

hroughout the field; however, on average, season-average ECfwnd ECss were 79% and 130% higher, respectively, in bottom basinsompared to the top basins (Table 3). Within a field, the maximumbserved ECfw tended to occur in the bottom basin between week

and week 7 after planting, and ranged from 0.36 to 6.06 dS m−1

Table 2). Spikes in ECfw early in the season tended to coincide withow field water depth (Fig. 3). There was no consistent temporal

Css trend among fields (data not shown). Additionally, while thereas a positive correlation (R2=0.40) between season-average ECfw

nd season-average ECss (Fig. 4), there is poor correlation whenCss > 3 dS m−1. Ta

ble

2O

verv

iew

of

Irri

gati

Fiel

dIrr

igat

iC

1

Surf

ace

C2

Gro

un

C3

Surf

ace

C4a

Gro

un

C5a

Surf

ace

C6

Surf

ace

C7a

Gro

un

G1

Surf

ace

G2

Surf

ace

B1

Surf

ace

Y1

Surf

ace

afi

eld

wit

h

42 M. Marcos et al. / Agricultural Water Management 195 (2018) 37–46

Table 3Relative increases of soil or field water salinity from the top of field (A1 plot) tobottom of the field (C3 plot), with the 95% confidence interval in parentheses.

Relative Increase from top (A1) to bottom (C3) (%)

ECfw 78.8 *** (46.2–111.5)ECss 130.0 ** (42.3–217.6)ECe 53.4 ** (17.1–89.8)

ECfw: field water salinity. ECss: soil solution salinity.ECe: preseason soil saturated paste salinity.*, **, ***, correspond to the 0.05, 0.01, and 0.001 significance level.

Table 4Parameter estimates for mixed-effects models for yield loss due to salinity. Fieldsite was a random effect. Only field water EC values greater than the threshold(0.88 dS m−1) were used for these models.

Mixed Effects Model Fixed Effects Estimate 95% ConfidenceInterval

Field water EC ModelIntercept 13656.0 *** 12278–14903ECfw −1971.8 ** −2756 − −1151Marginal R2 0.46AICc 533.97

Soil EC ModelIntercept 12475.2 *** 10844 − 14088ECe −334.6 -654–2Marginal R2 0.16AICc 537.5

Field water + Soil EC ModelIntercept 14349.4 *** 12897 − 15827ECfw −1941.5 *** −2662 − −1226ECe −154.5 −375 − 47Marginal R2 0.50

*

3

as0rpvosmwycanRnmla

3

ztsRtd

Table 5Model accuracy parameters (R2 and Root Mean Squared Error (RMSE)) betweenpredictions from Eq. (4) and ground-truth values. For Observed, the entire datasetwas fit to Eq (4). For Validation, the leave-one-field-out cross-validations were fitto Eq. (4). The season average model includes measurements from week 1 to week15 after planting. The early season model includes measurements from week 1 toweek 7 after planting. The late season model includes measurements from week 8to week 15 after planting.

Model Observed Validation

R2 RMSE R2 RMSE

Season Average 0.69 0.149 0.60 0.171

AICc 534.97

,**,***, correspond to the 0.05, 0.01, and 0.001 significance level.

.3. Yield results

Using a mixed-effects modeling approach, a significant neg-tive correlation was observed between yield and field wateralinity above 0.88 dS m−1, with a 95% confidence interval of.35–1.36 dS m−1 (Fig. 5). In contrast, the same approach did notesult in a significant breakpoint for ECss and ECe (data not shown),otentially due to the nature of the data. There was a gap in ECealues between 1.88 and 3.88 dS m−1 (Supplementary Fig. 2), andnly three ECss data points per field. Therefore, the effect that soilalinity may have on yield loss was further assessed by comparingodels that included and excluded ECe and ECfw terms (Table 4),hich helped to determine the importance of the two variables to

ield loss. The yield model with solely ECfw resulted in a signifi-ant regression coefficient of −1971.8 kg ha−1 per dS m−1 and had

marginal R2 of 0.46 (Table 4). The yield model with solely ECe didot produce a significant regression coefficient, and had a marginal2 of 0.16. The yield model with both ECfw and ECe produced a sig-ificant regression coefficient for ECfw but not for ECe, while thearginal R2 was 0.50. Additionally, the model with only ECfw had a

ower AICc value than the models with only ECe or with both ECfwnd ECe.

.4. Modeling field water salinity

A model was developed that predicts season-average ECfw inero-drainage fields based on incoming irrigation water salinity,he lateral and longitudinal distance from the inlet, and preseason

2

oil salinity, with R of 0.69 when evaluated against all data and2 of 0.60 for the LOFO cross-validation predictions (Table 5). Forhe season-average water salinity model, lateral and longitudinalistance together contributed to 82% of the explained variance in

Early Season 0.76 0.194 0.68 0.227Late Season 0.57 0.149 0.19 0.239

the model, while preseason soil salinity contributed 18% (Table 6).Model results for the early-season (week 1 to week 7 after plant-ing) indicate that preseason soil salinity had a larger effect andcontributed to 49% of the explained variance in the model, whilelateral and longitudinal distance contributed to the remaining 51%(Table 6). In contrast, preseason soil salinity did not have a signif-icant effect in the late-season (week 8 to week 15 after planting)(Table 6). The early-season model also performed better than thelate-season model, as the validation R2 for the early-season modelwas 0.68, compared to 0.19 for the late-season model (Table 5).Additionally, while fields with conventional-drainage were notincluded in the development of the models, in the early-seasonmodel, observed relative water salinity of fields with conventional-drainage was always lower than the predicted value (Fig. 6 ).

4. Discussion

4.1. Field water EC is the most sensitive salinity metric for riceyield

Soil and water salinity are both commonly used to determinewhether rice yield may be adversely affected by salinity (Ayers andWestcot 1985; Maas and Grattan 1999). In this study, there weretwo soil salinity metrics, preseason soil salinity (ECe), which camefrom samples collected at each plot from 0 to 15 cm below the soilsurface prior to flooding, and soil solution salinity (ECss), whichwere weekly pore-water measurements at 10 cm below the soilsurface. Though there was a good correlation (R2 = 0.89) betweenseason-average ECss and preseason ECe (Supplementary Fig. 3), thetwo soil salinity metrics were employed for different analyses. Toobserve the relationship between soil and field water salinity, ECsswas used, as there were measurements for this metric at the sametime and location as ECfw. To study the relationship with yield,ECe was used, as there were ECe measurements at all plot loca-tions. Additionally, growers are more likely to gather soil samplesbefore the season, when the field is dry, and this is therefore a moreappropriate metric for yield.

In this study, a positive correlation (R2 = 0.40) was observedbetween season-average ECfw and season-average ECss (Fig. 4).Similarly, Dickey and Nuss (2002) and Scardaci et al. (2002), foundwithin-season correlations of soil and water salinity, with R2 valuesranging from 0.52 to 0.70. Grattan et al. (2002) found season-average ECfw and postseason ECe to correlate with R2 = 0.88. Highcorrelations between soil and field water salinity led Grattan et al.(2002) and Dickey and Nuss (2002) to effectively equate soil salinity(averaged from 0 to 15 cm below the soil surface), with field watersalinity. The poor correlation between ECfw and ECss observed atEC values >3 ds m−1 (Fig. 4) questions this approach. At lower soil

sssalinity concentrations there seems to be good correlation betweensoil and field water salinity but at higher soil salinity concentra-tions there is no clear relationship. The poor correlation at higher

M.

Marcos

et al.

/ A

gricultural W

ater M

anagement

195 (2018)

37–46

43

Table 6Model coefficients and variable contributions to the models to predict relative water EC (field water salinity relative to the irrigation water salinity) in zero-drainage fields. These models were developed using data from all fields.The variable contributions to the explained variation are normalized to sum 100%. The season average model includes measurements from week 1 to week 15 after planting. The early season model includes measurements fromweek 1 to week 7 after planting. The late season model includes measurements from week 8 to week 15 after planting. The 95% confidence intervals of the estimates are in parentheses below each estimate.

Model Model Coefficients Variable Contribution to explained R2 (%)

Intercept Longitudinal Distance Lateral Distance Preseason ECe Longitudinal Distance Lateral Distance Preseason ECe

Season Average 9.293e-01 *** 5.057e-04 *** 9.994e-04 *** 4.887e-02 *** 68.9 13.4 17.7(8.31e-01–1.028) (3.998e-04 − 6.116e-04) (6.20e-04 − 1.379e-03) (2.884e-02 − 6.890e-02)

Early Season 8.668e-01 *** 6.196e-04 *** 1.051e-03 *** 1.225e-01 *** 46.6 4.4 49.0(7.380e-01 − 9.956e-01) (4.808e-04 − 7.585e-04) (5.492e-04 − 1.553e-03) (9.627e-02 − 1.487e-01)

Late Season 9.796e-01 *** 4.270e-04 *** 7.799e-04 *** -8.691e-03 79.1 19.2 1.7(8.809e-01 − 1.078) (3.204e-04 − 5.337e-04) (3.975e-04 − 1.162e-03) (-2.884e-02 − 1.146e-02)

*, **, ***, correspond to the 0.05, 0.01, and 0.001 significance level.

Fig. 6.

Relation

ship

s betw

een observed

relative w

ater EC

(field

water

salinity

relative to

irrigation w

ater salin

ity) an

d leave-on

e-field

-out

validation

pred

ictions

usin

g Eq.

4, for

season-average,

early-season,

and

late-season m

odels

in fi

elds

un

der

zero-drain

age (fi

lled d

ots). Field

s u

nd

er con

vention

al-drain

age w

ere n

ot in

clud

edin

the

develop

men

t of

the

mod

els; h

owever,

the

relationsh

ip betw

een observed

relative w

ater EC

and

pred

ictions

usin

g th

e fi

nal

mod

el are

show

n (u

nfi

lled d

ots).

4 ater M

sfi

aimsvi(gs

bt1mtfSsde

4p

bsas12gboiwssstr1dopw

0odewtiegB

Tip1T

4 M. Marcos et al. / Agricultural W

oil salinity necessitates differentiation between soil salinity andeld water salinity in flooded rice cropping systems.

Field water salinity correlated better with yield, and produced better model based on AICc values, than did preseason soil salin-ty (Table 4). Furthermore, adding preseason soil salinity to the

odel with field water salinity did not produce a significant regres-ion coefficient for soil salinity, worsened the model based on AICcalues, and did not greatly improve the models predictive capac-ty. These results, along with high yields observed under high ECeSupplementary Fig. 2) and high ECss (Supplementary Fig. 4), sug-est that ECfw is the most sensitive index when seeking to predictalinity induced yield loss.

Rice yield responding primarily to water salinity stress maye explained by the morphology of rice roots. Under flooded cul-ure, rice plants develop many surface roots (Morita and Yamazaki993), and the rooting depth is partially controlled by environ-ental conditions (Yoshida 1981). Previous studies have shown

hat under heterogeneous salinity conditions, roots take up waterrom the least saline areas (Bazihizina et al., 2012; Homaee andchmidhalter 2008). It is therefore likely that in fields where soilalinity is much higher than water salinity, rice roots preferentiallyeveloped and absorbed water from the surface, and thereby onlyxhibited a response to water salinity stress.

.2. Threshold concentration of field water salinity is lower thanreviously reported

The effect field water salinity will have on rice yield is a com-ination of the salt concentration, the length of exposure to thetress, and the plant susceptibility at that growth stage (Läuchlind Grattan 2007). While it is known that rice is very sensitive toalinity during tillering (Heenan et al., 1988; Pearson and Bernstein959; Zeng et al., 2001) and during reproduction (Castillo et al.,007; Heenan et al., 1988; Fraga et al., 2010), it is not clear whichrowth stage is most sensitive to salinity nor the relative contri-ution that salinity stress would have to yield loss if salinity stressccurred during multiple growth stages. Additionally, reductions

n stand establishment were not observed in this study. Therefore,ater salinity concentrations used for the yield threshold analy-

is were a season-average ECfw developed from weekly field wateralinity measurements averaged over the duration of the growingeason in which the field was flooded. This approach is similar tohat taken by others that have contributed to threshold estimates inice (Ehrler 1960; Grattan et al., 2002; Narale et al., 1969; Pearson959; Venkateswarlu et al., 1972). While this study was able toevelop a yield threshold for field water salinity, the narrow rangef season-average field water salinity observed (0.10–2.72 dS m−1),revents characterization of crop yield response to very high fieldater salinity.

The season-average ECfw yield threshold was estimated to be.88 dS m−1 (Fig. 5). This threshold is well below the previous reportf 1.9 dS m−1 (Grattan et al., 2002) and this may be due to theifferent conditions under which values were obtained. Grattant al. (2002) conducted their study under steady-state conditions,hereby salinity concentrations remained constant from seeding

o maturity. In contrast, in this study, field water salinity var-

ed during the growing season (Table 2; Fig. 3), with rice plantsxperiencing spikes in salinity during tillering, a very salt-sensitiverowth stage (Heenan et al., 1988; Fraga et al., 2010; Pearson andernstein 1959; Zeng et al., 2001). This resulted in a lower threshold

he early-season model includes measurements from week 1 to week 7 after plant-ng. The late-season model includes measurements from week 8 to week 15 afterlanting. The season-average model includes measurements from week 1 to week5 after planting. R2 and Root Mean Squared Error (RMSE) are reported in the figure.he dashed line is a 1:1 line, while the solid line is the line of best fit.

anagement 195 (2018) 37–46

estimate, as the season-average salinity concentration incorporateshigher field water salinity during a sensitive growth stage, andlower field water salinity during less sensitive growth stages. Ide-ally, growth-stage specific salinity thresholds could be developed;however, this was not feasible in this experiment, as the salinityconcentrations were not controlled.

Nevertheless, the approach taken to establish this thresh-old estimate addresses a key concern that many have with thetraditional threshold-slope approach. Generally, salinity studiesattempt to characterize the entire range of salinity stress, from0 to 100% yield loss; therefore, they tend to have too few obser-vations near the threshold value and too many above what isrealistic in grower’s fields. This tends to result in poor defini-tion of the yield threshold (van Genuchten and Hoffman, 1984).For the previous 1.9 dS m−1 threshold, the 95% confidence inter-val was 0.6–3.2 dS m−1 (Grattan et al., 2002; Grieve et al., 2012). Inthis study, the 95% confidence interval for the 0.88 dS m−1 thresh-old was 0.35–1.36 dS m−1 (Fig. 5). The improved definition of thisthreshold estimate is likely due to this study having more observa-tions near the expected threshold value. This, along with this studybeing done under commercial field conditions, with realistic sea-sonal variations in ECfw, makes this season-average threshold moreapplicable to commercial rice fields. Future work should be done todevelop salinity thresholds at various growth stages, as this couldnot be achieved in this study.

4.3. Spatio-temporal field water salinity dynamics

Quantifying the spatial and temporal variation of water salinityin rice fields is the first step towards developing potential man-agement solutions to prevent salinity induced yield loss. Among allfields, consistent spatial water salinity patterns emerged: season-average ECfw was greater in C3 positions than in A1 positions byan average of 79% (Table 3). Additionally, in most field sites, themaximum ECfw occurred in the bottom basin, and ranged from0.36 to 6.06 dS m−1 (Table 2). Similarly, Scardaci et al. (2002),Simmonds et al. (2013) and Shannon et al. (1998), all reportedhigher water salinity in bottom basins of rice fields. It is likelythat ECfw increases with increasing distance from the irrigationinlet due to evapo-concentration of salts in the field water and thesubsequent movement of that salt concentrated water down thefield.

A consistent temporal water salinity pattern also emerged: rela-tive ECfw is higher early in the season (Fig. 3, Fig. 6) and ECfw maximagenerally occurred from week 2 to week 7 after planting (Table 2).This is similar to results reported by Scardaci et al. (2002). The risein ECfw early in the season is likely due to two factors. First, lowcanopy cover during this period allows for high rates of evapora-tion from the water surface thereby concentrating salts in the fieldwater. Second, it is common for growers to allow field water tosubside early in the season for the application of foliar herbicides(University of California Cooperative Extension, 2015). As the fieldwater subsides, the surface area to volume ratio of the field waterincreases such that evapo-concentration increases. In this study,low water depths during this period coincided with higher watersalinity (Fig. 3).

4.4. Modeling field water salinity

Consistent spatial trends of water salinity among fields allowedfor the development of a model capable of predicting season-average water salinity throughout a rice field under zero-drainage

(Fig. 6; Table 6). Position within a field, which is a combination ofthe longitudinal distance and lateral distance from the irrigationinlet, accounted for 82% of the explained variation in the model.This result indicates the importance that distance, and in effect

ater M

fiFgc

ancmmbfloTsefiriui

ecmssdsia

dfdfiuisTiui

dTmeefit

4

wasegnIstwd

M. Marcos et al. / Agricultural W

eld size, has on water salinity build-up within flooded rice fields.or the same irrigation water and soil salinity, larger fields are atreater risk of having field water salinity increase to yield-reducingoncentrations.

The other factor influencing water salinity build-up within field was the preseason soil salinity. Soil salinity had a sig-ificant effect on water salinity build-up early in the season,ontributed to 49% of the explained variance in the early-seasonodel (Table 6). Furthermore, among all field sites, the higher ECfwaxima occurred in sites with high ECe (Table 2). These results may

e explained by considering vertical water and solute movement inooded rice fields. Rice fields in the region tend to have low ratesf percolation during the growing season (Linquist et al., 2015).his, along with low transpiration rates early in the growing sea-on results in low advective flow downward. Studies by Bachandt al. (2014a,b) show upward diffusion of solutes in flooded riceelds occurs during periods of minimal transpiration and that theate of diffusion is dictated by concentration gradients. Therefore,t is likely that in fields with high soil salinity, salts were diffusingpward into the field water early in the growing season, contribut-

ng to greater observed field water salinity.Temporal differences are also evident from these models. The

arly-season model had a substantially higher validation R2, 0.68,ompared to the late-season model, 0.19 (Fig. 6). This is due toodel parameters having less influence later in the growing sea-

on. Salts from the soil are less likely to diffuse upward later in theeason when transpiration is high and there is high advective flowownward. Additionally, low evaporation later in the growing sea-on, due to canopy cover, and a higher water depth, decreases thenfluence that the spatial position within a field (i.e. longitudinalnd lateral distance) has on water salinity build-up within a field.

These models also allow comparisons to be made between zero-rainage and conventional-drainage fields. While the sample size

or comparison was limited (only 3 fields under conventional-rainage), there is evidence suggesting tailwater drainage reduceseld water salinity. During the early-season, water salinity in fieldsnder conventional-drainage was overestimated (Fig. 6), indicat-

ng that tailwater drainage early in the season reduces field wateralinity. This is similar to results reported by Scardaci et al. (2002).ailwater drainage helps remove dissolved salts that accumulate

n field water. The tailwater drainage early in the season, in fieldsnder conventional-drainage, thus resulted in reduced water salin-

ty in those fields.These modeling results highlight the ability to sufficiently pre-

ict water salinity in rice fields with a few simple parameters.his study did not consider flow rates, slope of the field, or cli-atic effects. These factors could have an important contribution to

xplaining water salinity in a field and may have improved modelstimates. However, it is shown, that with relatively few factors,eld water salinity can be reasonably well predicted which facili-

ates application of these results for use by growers.

.5. Management implications

This study has several important implications for managingater to control salinity in rice fields. First, in fields with low soil

nd irrigation water salinity, practicing zero-drainage is a viabletrategy to reduce water usage without harming yield. Seven of theight field sites studied under zero-drainage, had low soil and irri-ation water salinity, and in those sites, the field-average ECfw didot increase beyond the 0.88 dS m−1 salinity threshold (Table 2).

n the long-term however, zero-drainage may result in secondary

oil salinization if amelioration efforts are not employed. One wayo reduce the potential long-term soil salinity build-up is throughinter flooding, a common practice in California to aid in rice straw

ecomposition (Linquist et al., 2006). Winter flooding can bring

anagement 195 (2018) 37–46 45

up and export solutes that have built-up during the growing sea-son (Bachand et al., 2014a). The degree to which winter floodingcan reduce salt concentrations in rice fields under long-term zero-drainage management merits further investigation.

Second, in fields with either high soil or irrigation water salin-ity, maintaining a higher water depth and allowing for tailwaterdrainage early in the season will help reduce field water salinity.The highest field water salinity occurred early in the growing sea-son (Table 2; Fig. 3) when rice is sensitive to salinity (Heenan et al.,1988; Fraga et al., 2010; Pearson and Bernstein 1959; Zeng et al.,2001). In these high salinity fields, salt flow from the soil to thefield water and high rates of evapo-concentration, may increasefield water salinity to yield-reducing concentrations. Increasingwater depth during this period allows for dilution of dissolvedsalts. Additionally, allowing tailwater drainage during this periodmay also help reduce field water salinity (Fig. 6). These practices,however, may not be feasible with many pesticide regimes com-monly employed in California, as contact herbicides require lowwater depth to ensure proper coverage, and water holding periodsimposed after pesticide applications prevents tailwater drainage.

Lastly, in most fields studied, season-average field water salinityreached concentrations that were 50% greater than the irrigationwater salinity (Fig. 6). Therefore, to avoid season-average fieldwater salinity increasing beyond the 0.88 dS m−1 threshold in thebottom of an average sized field (35 ha in this study), it is recom-mended to maintain irrigation water salinity below 0.6 dS m−1. Thiscriterion for irrigation water salinity is similar to the 0.75 dS m−1

reported by Finfrock et al. (1960). To maintain irrigation watersalinity below 0.6 dS m−1, the use of groundwater as an irrigationsource should be minimized, and irrigation districts should care-fully portion the amount of high salinity recycled water allowedin the supplied surface water. Furthermore, as increases in watersalinity within a field are largely a function of distance from theirrigation inlet, in areas with high soil or irrigation water salinity,smaller fields and multiple irrigation inlets should be considered.

Acknowledgements

We would like to thank the California Rice Research Board forproviding funding for this study. Furthermore, we would like tothank the staff and students in the Agroecosystems laboratory atUC Davis, most especially Cesar Abrenilla, for their support andassistance throughout this project.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at https://doi.org/10.1016/j.agwat.2017.09.016.

References

Ayers, R.S., Westcot, D.W., 1985. Water quality for agriculture. FAO Irrig. Drain, Vol.29. FAO, Rome.

Bachand, P.A.M., Bachand, S.M., Fleck, J.A., Alpers, C.N., Stephenson, M.,Windham-Myers, L., 2014a. Methylmercury production in and export fromagricultural wetlands in california: USA: the need to account for physicaltransport processes into and out of the root zone. Sci. Total Environ. 472,957–970.

Bachand, P.A.M., Bachand, S.M., Fleck, J.A., Anderson, F., Windham-Myers, L.,2014b. Differentiating transpiration from evaporation in seasonal agriculturalwetland and the link to advective fluxes in the root zone. Sci. Total Environ.484, 232–248.

Bartoń, K., 2016. MuMIn: Multi-Model Inference. https://CRAN.R-project.org/package=MuMIn.

Bates, D., Maechler, M., Bolker, B., Walker, S., 2016. Lme4: Linear Mixed-Effects

Models Using Eigen and S4. http://CRAN.R-project.org/package=lme4.

Bazihizina, N., Barrett-Lennard, E.G., Colmer, T.D., 2012. Plant growth andphysiology under heterogeneous salinity. Plant Soil 354, 1–19.

Becker, R.A., Wilks, A.R., Brownrigg, R., Minka, T.P., Deckmyn, A., 2016. Maps: DrawGeographical Maps. https://CRAN. R-project. org/package= maps.

http://dx.doi.org/10.1016/j.agwat.2017.09.016http://dx.doi.org/10.1016/j.agwat.2017.09.016http://dx.doi.org/10.1016/j.agwat.2017.09.016http://dx.doi.org/10.1016/j.agwat.2017.09.016http://dx.doi.org/10.1016/j.agwat.2017.09.016http://dx.doi.org/10.1016/j.agwat.2017.09.016http://dx.doi.org/10.1016/j.agwat.2017.09.016http://dx.doi.org/10.1016/j.agwat.2017.09.016http://dx.doi.org/10.1016/j.agwat.2017.09.016http://dx.doi.org/10.1016/j.agwat.2017.09.016http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0005http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0005http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0005http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0005http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0005http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0005http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0005http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0005http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0005http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0005http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0005http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0010http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0010http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0010http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0010http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0010http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0010http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0010http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0010http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0010http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0010http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0010http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0010http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0010http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0010http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0010http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0010http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0010http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0010http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0010http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0010http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0010http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0010http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0010http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0010http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0010http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0010http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0010http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0010http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0010http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0010http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0010http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0010http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0010http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0015http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0015http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0015http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0015http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0015http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0015http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0015http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0015http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0015http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0015http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0015http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0015http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0015http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0015http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0015http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0015http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0015http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0015http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0015http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0015http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0015http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0015http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0015http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0015http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0015https://CRAN.R-project.org/package=MuMInhttps://CRAN.R-project.org/package=MuMInhttps://CRAN.R-project.org/package=MuMInhttps://CRAN.R-project.org/package=MuMInhttps://CRAN.R-project.org/package=MuMInhttps://CRAN.R-project.org/package=MuMInhttp://CRAN.R-project.org/package=lme4http://CRAN.R-project.org/package=lme4http://CRAN.R-project.org/package=lme4http://CRAN.R-project.org/package=lme4http://CRAN.R-project.org/package=lme4http://CRAN.R-project.org/package=lme4http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0030http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0030http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0030http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0030http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0030http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0030http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0030http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0030http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0030http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0030http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0030http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0030http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0030https://CRAN. R-project. org/package= mapshttps://CRAN. R-project. org/package= mapshttps://CRAN. R-project. org/package= mapshttps://CRAN. R-project. org/package= mapshttps://CRAN. R-project. org/package= mapshttps://CRAN. R-project. org/package= maps

4 ater M

B

B

B

B

C

C

C

D

EF

F

F

G

G

G

G

H

H

H

H

H

H

H

K

K

K

L

L

L

L

L

Zeng, L., Shannon, M.C., Lesch, S.M., 2001. Timing of salinity stress affects ricegrowth and yield components. Agric. Water Manage. 48, 191–206.

van Genuchten, M.Th., Hoffman, G.J., 1984. Analysis of crop salt tolerance data. In:Shainberg, I., Shalhevet, J. (Eds.), Soil Salinity Under Irrigation, Processes andManagement: Ecological Studies 51. Springer-Verlag, New York, pp. 258–271.

6 M. Marcos et al. / Agricultural W

ivand, R., Lewin-Koh, N., 2016. Maptools: Tools for Reading and Handling SpatialObjects. https://CRAN. R-project. org/package= maptools.

ouman, B.A.M., Humphreys, E., Tuong, T.P., Barker, R., 2007. Rice Water Adv.Agron. 92, 187–237.

urnham, K.P., Anderson, D.R., 2002. . In: Model Selection andMultimodel Inference: A Practical Information-theoretic Approach, 2ndedition. Springer-Verlag, New York.

urt, R., 2014. Soil Survey Laboratory Methods Manual. SSIR No. 42. GovernmentPrinting Office, Washington, DC: U.S (Version 5.0. NRCS-USDA).

IMIS, 2016. California Irrigation Management Information System (InternetResource) http://www.cimis.water.ca.gov/WSNReportCriteria.aspx.

astillo, E.G., Tuong, T.P., Ismail, A.M., Inubushi, K., 2007. Response to salinity inrice: comparative effects of osmotic and ionic stresses. Plant Prod. Sci. 10,159–170.

onnor, J.D., Schwabe, K., King, D., Knapp, K., 2012. Irrigated agriculture andclimate change: the influence of water supply variability and salinity onadaptation. Ecol. Econ. 77, 149–157.

ickey, J., Nuss, G., 2002. Salinity distribution and impact in the sacramento valley.USCID Water Management Conference on Helping Irrigating Agriculture Adjustto TMDLs.

hrler, W., 1960. Some effects of salinity on rice. Bot. Gaz. 122, 102–104.infrock, D.C., Raney, F.C., Miller, M.D., Booher, L.J., 1960. Water management in

rice production. In: Division of Agricultural Sciences. University of California,pp. 131 (Leaf).

raga, T.I., Carmona, F.C., Anghinoni, I., Junior, S.A.G., Marcolin, E., 2010. Floodedrice yield as affected by levels of water salinity in different stages of its cycle. R.Bras. Ci Solo 34, 175–182.

raiture, C.D., Wichelns, D., 2010. Satisfying future water demands for agriculture.Agric. Water Manage. 97, 502–511.

rattan, S.R., Zeng, L., Shannon, M.C., Roberts, S.R., 2002. Rice is more sensitive tosalinity than previously thought. Calif. Agric. 56, 189–198.

rattan, S.R., 2002. Irrigation Water Salinity and Crop Production. Water QualityFact Sheet. Division of Agriculture and Natural Resources. University ofCalifornia. UC DANR electronic publication no. 8066, Oakland, CA.

rieve, C.M., Grattan, S.R., Maas, E.V., 2012. Plant salt tolerance. In: Wallender,W.W., Tanji, K.K. (Eds.), ASCE Manual and Reports on Engineering Practice No.71 Agricultural Salinity Assessment and Management. , 2nd edition. ASCE,Reston, VA, pp. 405–459 (Chapter 13).

roemping, U., Matthias, L., 2013. h. In: Relaimpo: Relative Importance ofRegressors in Linear Models. ttps://CRAN.R-project.org/package=relaimpo.

anak, E., Lund, J.R., 2012. Adapting California’s water management to climatechange. Clim. Change 111, 17–44.

eenan, D.P., Lewin, L.G., McCaffery, D.W., 1988. Salinity tolerance in rice varietiesat different growth stages. Aust. J. Exp. Agric. 28, 343–349.

ijmans, R.J., 2015. Raster: Geographic Data Analysis and Modeling. https://CRAN.R-project.org/package=raster.

ill, J.E., Scardaci, S.C., Roberts, S.R., Tiedeman, J., Williams, J.F., 1991. Rice irrigationsystems for tailwater management. Univ. Calif. Div. Ag. Nat. Res. Pub. 21490.

ill, J.E., Williams, J.F., Mutters, R.G., Greer, C.A., 2006. The California rice croppingsystem: agronomic and natural resource issues for long-term sustainability.Paddy Water Environ. 4, 13–19.

omaee, M., Schmidhalter, U., 2008. Water integration by plants root undernon-uniform soil salinity. Irrig. Sci. 27, 83–95.

owitt, R.E., MacEwan, D., Medellín-Azuara, J., Lund, J.R., Sumner, D.A., 2015.Economic Analysis of the 2015 Drought for California Agriculture. Center forWatershed Sciences, UC, Davis, Davis, CA.

ijne, J.W., 2006. Abiotic stress and water scarcity: identifying and resolvingconflicts from plant level to global level. Field Crop Res. 97, 3–18.

uhn, M., 2016. Caret: Classification and Regression Training. https://cran.r-project.org/package=caret.

uznetsova, A., Brockhoff, P.B., Christensen, R.H.B., 2016. LmerTest: Tests in LinearMixed Effects Models. https://cran.r-project.org/package=lmerTest.

äuchli, A., Grattan, S.R., 2007. Plant growth and development under salinity stress.In: Jenks, M.A., Hasegawa, P.M., Mohan, J.S. (Eds.), Advances in MolecularBreeding Towards Drought and Salt Tolerant Crops. Springer, Berlin, pp. 1–32.

ekakis, E., Aschonitis, V., Pavlatou-Ve, A., Papadopoulos, A., Antonopoulos, V.,2015. Analysis of temporal variation of soil salinity during the growing seasonin a flooded rice field of thessaloniki plain-Greece. Agron. J. 5, 35–54.

indeman, R.H., Merenda, P.F., Gold, R.Z., 1980. Introduction to Bivariate andMultivariate Analysis. Scott Foresman, Glenview, IL.

inquist, B.A., Brouder, S.M., Hill, J.E., 2006. Winter straw and water managementeffects on soil nitrogen dynamics in California rice systems. Agron. J. 98,

1050–1059.

inquist, B.A., Snyder, R., Anderson, F., Espino, L., Inglese, G., Marras, S., Moratiel, R.,Mutters, R.G., Nicolosi, P., Rejmanek, H., Russo, A., Shapland, T., Song, Z.,Swelam, A., Tindula, G., Hill, J.E., 2015. Water balances and evapotranspirationin water- and dry-seeded rice systems. Irrig. Sci. 33, 375–385.

anagement 195 (2018) 37–46

Maas, E.V., Grattan, S.R., 1999. Crop yields as affected by salinity. In: Skaggs, R.W.,van Schilfgaarde, J. (Eds.), Agricultural Drainage. Madison. ASA-CSSA-SSSA, WI,pp. 55–108.

Maas, E.V., Hoffman, G.J., 1977. Crop salt tolerance—current assessment. J. Irrig.Drain. Eng. 103, 115–134.

Mirchi, A., Madani, K., Roos, M., Watkins, D.W., 2013. Climate change impacts oncalifornia’s water resources. In: Schwabe, K., et al. (Eds.), Drought in Arid andSemi-Arid Regions. , pp. 301–322 (Chapter 17).

Molden, D., Oweis, T., Steduto, P., Bindraban, P., Hanjra, M.A., Kijne, J., 2010.Improving agricultural water productivity: between optimism and caution.Agric. Water. Manage. 97, 528–535.

Morita, S., Yamazaki, K., 1993. Root system. In: Matsuo, T., Hoshikawa, K. (Eds.),Science of the Rice Plant, Vol. 1. Morphology. Food and Agriculture PolicyResearch Center, Tokyo, pp. 161–186.

Munns, R., Tester, M., 2008. Mechanisms of salinity tolerance. Annu. Rev. Plant Biol.59, 651–681.

Nakagawa, S., Schielzeth, H., 2013. A general and simple method for obtaining R2from Generalized Linear Mixed-effects Models. Methods Ecol. Evol. 4, 133–142.

Narale, R.P., Subramanyam, T.K., Mukherjee, R.K., 1969. Influence of salinity ongermination, vegetative growth, and grain yield of rice (Oryza sativa var.Dular). Agron. J. 61, 311–344.

Pearson, G.A., Bernstein, L., 1959. Salinity effects at several growth stages of rice.Agron. J. 51, 654–658.

Pearson, G.A., 1959. Factors influencing salinity of submerged soils and growth ofCaloro rice. Soil Sci. 87, 198–206.

Pike, A.C., Mueller, T.G., Schörgendorfer, A., Shearer, S.A., Karathanasis, A.D., 2009.Erosion index derived from terrain attributes using logistic regression andneural networks. Agron. J. 101, 1068–1079.

R Core Team, 2015. R: A Language and Environment for Statistical Computing. RFoundation for Statistical Computing, Vienna, Austria https://www.R-project.org/.

Scardaci, S.C., Eke, A.U., Hill, J.E., Shannon, M.C., Rhoades, J.D., 1996. Water and soilsalinity studies on california rice. In: Rice. University of California CooperativeExtension (Publication No. 2).

Scardaci, S.C., Shannon, M.C., Grattan, S.R., Eke, A.U., Roberts, S.R., Goldman-Smith,S., Hill, J.E., 2002. Water management practices can affect salinity in rice fields.Calif. Agric. 56, 184–188.

Schewe, J., Heinke, J., Gerten, D., Haddeland, I., Arnell, N.W., Clark, D.B., Dankers, R.,Eisner, S., Fekete, B.M., Colon-Gonzalez, F.J., Gosling, S.N., Kim, H., Liu, X.,Masaki, Y., Portmann, F.T., Satoh, Y., Stacke, T., Tang, Q., Wada, Y., Wisser, D.,Albrecht, T., Frieler, K., Piontek, F., Warszawski, L., Kabat, P., 2014. Multimodalassessment of water scarcity under climate change. PNAS 111, 3245–3250.

Scudiero, E., Skaggs, T.H., Corwin, D.L., 2015. Regional-scale soil salinity assessmentusing Landsat ETM + canopy reflectance. Remote Sens. Environ. 169, 335–343.

Shalhevet, J., 1994. Using water of marginal quality for crop production: majorissues. Agric. Water Manage. 25, 233–269.

Shannon, M.C., Rhoades, J.D., Draper, J.H., Scardaci, S.C., Spyres, M.D., 1998.Assessment of salt tolerance in rice cultivars in response to salinity problemsin california. Crop Sci. 38, 394–398.

Simmonds, M.B., Plant, R.E., Peña-Barragán, J.M., Van Kessel, C., Hill, J.E., Linquist,B.A., 2013. Underlying causes of yield spatial variability and potential forprecision management in rice systems. Precis. Agric. 14, 512–540.

Stevens, A., Miralles, I., Van Wesemael, B., 2012. Soil organic carbon predictions byairborne imaging spectroscopy: comparing cross-Validation and validation.Soil Sci. Soc. Am. J. 76, 2174–2183.

USDA, 2013. Farm and Ranch Irrigation Survey (FRIS) (Table 36.).University of California Cooperative Extension, 2015. California Rice Production

Workshop Manual.Venkateswarlu, J., Ramesam, M., Mohan, Murali, Rao, G.V., 1972. Salt tolerance in

rice varieties. J. Ind. Soc. Soil Sci. 20, 169–213.Wallace, J.S., 2000. Increasing agricultural water use efficiency to meet future food

production. Agric. Ecosyst. Environ. 82, 105–119.Wickham, H., Chang, W., 2016. Ggplot2: An Implementation of the Grammar of

Graphics. https://cran.r-project.org/package=ggplot2.Wickham, H., 2016. Gtable: Arrange ‘Grobs’ in Tables. https://cran.r-project.org/

package=gtable.Yoshida, S., 1981. Fundamental of Rice Crop Science. IRRI, Los Baños Philippines.

https://CRAN. R-project. org/package= maptoolshttps://CRAN. R-project. org/package= maptoolshttps://CRAN. R-project. org/package= maptoolshttps://CRAN. R-project. org/package= maptoolshttps://CRAN. R-project. org/package= maptoolshttps://CRAN. R-project. org/package= maptoolshttp://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0045http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0045http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0045http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0045http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0045http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0045http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0045http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0045http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0050http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0050http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0050http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0050http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0050http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0050http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0050http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0050http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0050http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0050http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0050http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0050http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0050http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0050http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0050http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0050http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0055http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0055http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0055http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0055http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0055http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0055http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0055http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0055http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0055http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0055http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0055http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0055http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0055http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0055http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0055http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0055http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0055http://www.cimis.water.ca.gov/WSNReportCriteria.aspxhttp://www.cimis.water.ca.gov/WSNReportCriteria.aspxhttp://www.cimis.water.ca.gov/WSNReportCriteria.aspxhttp://www.cimis.water.ca.gov/WSNReportCriteria.aspxhttp://www.cimis.water.ca.gov/WSNReportCriteria.aspxhttp://www.cimis.water.ca.gov/WSNReportCriteria.aspxhttp://www.cimis.water.ca.gov/WSNReportCriteria.aspxhttp://www.cimis.water.ca.gov/WSNReportCriteria.aspxhttp://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0065http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0065http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0065http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0065http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0065http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0065http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0065http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0065http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0065http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0065http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0065http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0065http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0065http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0065http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0065http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0065http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0065http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0065http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0065http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0070http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0070http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0070http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0070http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0070http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0070http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0070http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0070http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0070http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0070http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0070http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0070http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0070http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0070http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0070http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0070http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0070http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0070http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0070http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0070http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0070http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0075http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0075http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0075http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0075http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0075http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0075http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0075http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0075http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0075http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0075http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0075http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0075http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0075http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0075http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0075http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0075http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0075http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0075http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0075http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0080http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0080http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0080http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0080http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0080http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0080http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0080http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0080http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0080http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0080http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0080http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0080http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0085http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0085http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0085http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0085http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0085http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0085http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0085http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0085http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0085http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0085http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0085http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0085http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0085http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0085http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0085http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0085http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0090http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0090http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0090http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0090http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0090http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0090http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0090http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0090http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0090http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0090http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0090http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0090http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0090http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0090http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0090http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0090http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0090http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0090http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0090http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0090http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0090http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0090http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0090http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0090http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0095http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0095http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0095http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0095http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0095http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0095http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0095http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0095http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0095http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0095http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0095http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0095http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0095http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0100http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0100http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0100http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0100http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0100http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0100http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0100http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0100http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0100http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0100http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0100http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0100http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0100http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0100http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0100http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0105http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0105http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0105http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0105http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0105http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0105http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0105http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0105http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0105http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0105http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0105http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0105http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0105http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0105http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0105http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0105http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0105http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0105http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0105http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0105http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0105http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0105http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0105http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0105http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0105http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0105http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0105http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0110http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0110http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0110http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0110http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0110http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0110http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0110http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0110http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0110http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0110http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0110http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0110http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0110http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0110http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0110http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0110http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0110http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0110http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0110http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0110http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0110http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0110http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0110http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0110http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0110http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0110http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0110http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0110http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0110http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0110http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0110http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0110http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0110http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0110http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0110http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0115http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0115http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0115http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0115http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0115http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0115http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0115http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0115http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0115http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0115http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0115http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0120http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0120http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0120http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0120http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0120http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0120http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0120http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0120http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0120http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0120http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0120http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0120http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0120http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0125http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0125http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0125http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0125http://refhub.elsevier.com/S0378-3774(17)30311-6/sbref0125http://refhub.elsevier.com/S0378-3774(17)303